Abstract: Quantum entanglement is a fundamental resource enabling a wide range of applications in quantum communication and computation. Reliable inference of entanglement from experimental data is constrained by noise, destructive measurements, and limited copy budgets, rendering full state tomography statistically and operationally inefficient for decision-oriented tasks. This seminar presents two complementary, provably grounded approaches to entanglement inference that operate at the level of measurement statistics. In particular, we consider the finite-sample estimation of operational entanglement functionals—such as concurrence and related quality parameters—from observables implementable via LOCC measurement circuits, and use these estimates to determine which states satisfy a prescribed application-specific quality threshold. Casting this task as a thresholding bandit problem enables adaptive measurement allocation and yields instance-dependent guarantees on correctness and copy complexity. Second, we address the problem of producing rigorous entanglement certificates directly from measurement data. We develop a statistically valid certification pipeline that combines witness-based measurements, estimation from finite samples, and semidefinite programming duality to output explicit entanglement witnesses. Numerical experiments investigate detection difficulty based on geometric proximity to the separable boundary and alignment with the measurement geometry, revealing strong instance-dependent behaviour and empirically showing that this approach requires far fewer copies than full state tomography.

Event Details
Title: Quantum-Classical Co-design for Entanglement Inference: Quality Estimation and Certification from Measurement Data
Date: February 16, 2026 at 03:00 PM
Venue: Google meet (https://meet.google.com/chd-mqug-iyc)
Speaker: Ms. Bharati K (EE20D700)
Guide: Dr. Krishna Jagannathan
Type: PHD seminar

Updated: